RE: Problem with statistics

I downloaded your script and in the readme file I see that it is written according the recommendations for Informix Dynamic Server 7.31. Are they the same for IDS 7.3 ? -----Original Message----- From: Obnoxio The Clown [mailto:obnoxio@hotmail.com] Sent: dinsdag 28 maart 2000 13:03 To: doris.willaert@suzuki.be; informix-list@iiug.org Subject: RE: Problem with statistics From: Willaert Doris <doris.willaert@suzuki.be> > >The batch program does updates on one of the tables mentioned. This table >contains =/- 1.600.000 records and the program changes approximately 2000 - >3000 records (including index values). This should not cause any change to the index access paths, unless it includes an update statistics somewhere within the code...? And even then, I'd be very sceptical. >What does your script ? Does it perform update statistics high or low ? Get it and try it. It does a "proper", by the book US. >If I use only selects that follow index path, isn't a simple update >statistics enough and why not ? No. The optimiser doesn't just use the table size, it also considers the distribution of the data. Get the script, and try it tonight. Don't do your update stats in the morning, and see what happens. ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com

Willaert Doris wrote: > I downloaded your script and in the readme file I see that it is written > according the recommendations for Informix Dynamic Server 7.31. Are they the > same for IDS 7.3 ? > The recommendation were first published as part of the release notes to IDS 7.21 and have been moved to the Performance Guide for 7.3x and 9.2x and revised over time. The basic recommendation is the same as it first was with some refinements to increase the range of statements that the recommended suite of UPDATE STATISTICS statements will help. The recommendations are a How-To for creating the most useful set of stats in the shortest possible time. The BEST level of stats is accomplished by doing an individual UPDATE STATISTICS HIGH on EVERY column in every table (with the DISTRIBUTIONS ONLY clause) plus an UPDATE STATISTICS LOW on the key list of ever index. However, this takesa HUGE amount of time and resources so the recommendations were created. Art S. Kagel > > -----Original Message----- > From: Obnoxio The Clown [mailto:obnoxio@hotmail.com] > Sent: dinsdag 28 maart 2000 13:03 > To: doris.willaert@suzuki.be; informix-list@iiug.org > Subject: RE: Problem with statistics > > From: Willaert Doris <doris.willaert@suzuki.be> > > > >The batch program does updates on one of the tables mentioned. This table > >contains =/- 1.600.000 records and the program changes approximately 2000 - > >3000 records (including index values). > > This should not cause any change to the index access paths, unless it > includes an update statistics somewhere within the code...? And even then, > I'd be very sceptical. > > >What does your script ? Does it perform update statistics high or low ? > > Get it and try it. It does a "proper", by the book US. > > >If I use only selects that follow index path, isn't a simple update > >statistics enough and why not ? > > No. The optimiser doesn't just use the table size, it also considers the > distribution of the data. Get the script, and try it tonight. Don't do your > update stats in the morning, and see what happens. > ______________________________________________________ > Get Your Private, Free Email at http://www.hotmail.com